Gait recognition method and device, and robot

Abstract

The disclosure provides a gait recognition method and device, and a robot, and belongs to the technical field of biometric identification. The method comprises the following steps: inputting first sensor data into a preset first deep learning model to obtain scores corresponding to walking scenes of each category. In response to the existence of a similar walking scene of a target walking scene, a score difference between the target walking scene and the similar walking scene is determined. In response to the score difference being less than a preset difference, the first sensor data and second sensor data are subtracted to obtain third sensor data. The third sensor data is input into a preset second deep learning model to obtain a walking scene of a target person. The gait recognition method and device, and the robot provided by the disclosure can identify the walking scene in which the target person is located, improve the accuracy of robot gait following, and help the robot to realize intelligent rehabilitation training.

Description

Technical Field

[0001] This disclosure belongs to the field of biometric technology, and more specifically, relates to a gait recognition method and device, and a robot. Background Technology

[0002] With the development of robotics technology, robots are increasingly being used in the field of human rehabilitation training. In this process, gait recognition methods have become crucial. Current robots can walk bipedally like humans and coordinate movements with the human body. When robots assist individuals in rehabilitation training, it is necessary to acquire real-time plantar pressure data throughout the entire gait cycle to adjust the individual's forward stride. However, existing robots generally suffer from low gait recognition accuracy and inefficiency. Therefore, improving the intelligence of robots and achieving standardized and efficient rehabilitation training has become a primary task. Summary of the Invention

[0003] The purpose of this disclosure is to provide a gait recognition method, device, and robot to improve the training effect of rehabilitation robots.

[0004] A first aspect of this disclosure provides a gait recognition method, including: The first sensor data is input into a pre-set first deep learning model to obtain scores corresponding to each category of walking scenarios. The first sensor data is the data collected by sensors installed on the robot during one gait cycle when the target person is walking; the target person is the person wearing the robot.

[0005] In response to the existence of similar walking scenarios in the target walking scenario, the score difference between the target walking scenario and the similar walking scenarios is determined. The target walking scenario is the walking scenario with the highest score in each category.

[0006] If the score difference is less than a preset difference, the difference between the first sensor data and the second sensor data is calculated to obtain the third sensor data. The third sensor data is then input into a preset second deep learning model to obtain the walking scene of the target person.

[0007] The second sensor data is the data collected by the sensor when the target person is stationary. The first deep learning model is trained based on sensor data samples in multiple walking scenarios, and the second deep learning model is trained based on sensor data samples in the target walking scenario and similar walking scenarios.

[0008] A second aspect of this disclosure provides a gait recognition device, comprising: The data input module is used to input the first sensor data into a preset first deep learning model to obtain scores corresponding to each category of walking scenarios. The first sensor data consists of data collected by sensors mounted on the robot during one gait cycle when a target person is walking; the target person is the person wearing the robot.

[0009] The score prediction module is used to determine the score difference between the target walking scene and similar walking scenes in response to the existence of similar walking scenes. The target walking scene is the walking scene with the highest score in each category.

[0010] The scene determination module, in response to a score difference less than a preset difference, calculates the difference between the first sensor data and the second sensor data to obtain third sensor data. This third sensor data is then input into a preset second deep learning model to obtain the walking scene of the target person.

[0011] The second sensor data is the data collected by the sensor when the target person is stationary. The first deep learning model is trained based on sensor data samples in multiple walking scenarios, and the second deep learning model is trained based on sensor data samples in the target walking scenario and similar walking scenarios.

[0012] A third aspect of this disclosure provides a robot including a processing terminal and a plurality of sensors connected to the processing terminal; the processing terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the gait recognition method described above.

[0013] A fourth aspect of this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the gait recognition method described above.

[0014] The beneficial effects of the gait recognition method, apparatus, and robot provided in this disclosure are as follows: This disclosure provides a gait recognition method for robots. Multiple different sensors are configured to collect data from a target person during a gait cycle and when stationary, resulting in first sensor data and second sensor data. A first deep learning model identifies and detects the first sensor data, obtaining scores for different walking scenarios. The walking scenario with the highest score is the target walking scenario. If similar walking scenarios exist, the difference between the first and second sensor data is calculated to obtain third sensor data. A second deep learning model then identifies and detects the third sensor data, thereby more accurately determining the target person's walking scenario. This method can identify similar gaits of a target person in different walking scenarios, effectively improving the accuracy of gait recognition, helping robots to accurately follow gait patterns, and enhancing the robot's training performance. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic flowchart of a gait recognition method provided in an embodiment of this disclosure; Figure 2 This is a structural block diagram of a gait recognition device provided in an embodiment of the present disclosure; Figure 3 This is a schematic block diagram of a processing terminal provided in an embodiment of the present disclosure. Detailed Implementation

[0017] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, so as to provide a thorough understanding of the embodiments of this disclosure. However, those skilled in the art will understand that this disclosure may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this disclosure with unnecessary detail.

[0018] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0019] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a gait recognition method provided in an embodiment of the present disclosure. The method includes: S101: Input the first sensor data into the preset first deep learning model to obtain the score corresponding to each category of walking scene. The first sensor data is the data of one gait cycle collected by the sensor set on the robot when the target person is walking. The target person is the person wearing the robot.

[0020] In this embodiment, a robot is provided. Driven by a series of elastic actuators, the robot uses flexible drive joints, exhibiting flexibility and reversibility. It can directly contact the human body, providing joint assistance as needed and in a timely manner, thus structurally ensuring the safety of human-robot interaction. The person wearing the robot is the target individual. Multiple sensors are installed on the robot to form an integrated multi-source sensor system, enabling real-time acquisition of the target individual's limb data and identification of their movement intentions.

[0021] The robot assists individuals in rehabilitation training across various walking scenarios. These scenarios can be customized to suit different environments. For example, this embodiment categorizes walking scenarios as flat ground, uphill, downhill, upstairs, and downstairs. The robot is equipped with multiple sensors that collect data from one gait cycle of the individual; this data constitutes the first sensor data. Based on sensor data samples from multiple walking scenarios, a deep learning model is pre-trained to predict scores for different walking scenarios; this model is the first deep learning model. The first sensor data is input into this model, and the model detects and identifies the sensor data to obtain scores for each category of walking scenario.

[0022] S102: In response to the existence of similar walking scenarios in the target walking scenario, determine the score difference between the target walking scenario and the similar walking scenarios. The target walking scenario is the walking scenario with the highest score in each category.

[0023] In this embodiment, based on the analysis of the scores corresponding to each category of walking scenarios, the walking scenario with the highest score in each category can be obtained, which is the target walking scenario. A judgment is made as to whether there are similar walking scenarios for the target walking scenario. If the judgment result indicates that there are similar walking scenarios for the target walking scenario, the score difference between the target walking scenario and its corresponding similar walking scenario is calculated. This embodiment provides a reference example: the target walking scenario predicted by the first deep learning model is uphill, with a score of 85 points. The similar walking scenario corresponding to this target walking scenario is upstairs, with a score of 80 points. The score difference between the target walking scenario and its corresponding similar walking scenario is 5 points. By analyzing the score difference between the target walking scenario and its corresponding similar walking scenario, it can be determined whether further walking scenario prediction is needed.

[0024] S103: In response to the score difference being less than a preset difference, the difference between the first sensor data and the second sensor data is calculated to obtain the third sensor data. The third sensor data is then input into a preset second deep learning model to obtain the walking scene of the target person.

[0025] The second sensor data is the data collected by the sensor when the target person is stationary. The first deep learning model is trained based on sensor data samples in multiple walking scenarios, and the second deep learning model is trained based on sensor data samples in the target walking scenario and similar walking scenarios.

[0026] In this embodiment, based on multiple sensors configured on the robot, limb data of the target person when in a stationary state is collected to obtain second sensor data. A preset difference is set based on the score difference between the target walking scene and its corresponding similar walking scenes. When the score difference between the target walking scene and its corresponding similar walking scenes is less than this preset difference, the first sensor data and the second sensor data are subtracted to obtain third sensor data. Based on the sensor data samples from the target walking scene and similar walking scenes, a deep learning model is pre-trained to further detect and identify the walking scene of the target person. This model is the second deep learning model. The third sensor data is input into the preset second deep learning model to determine the walking scene of the target person.

[0027] This embodiment provides a reference example. For instance, a preset difference value of 10 points is set. As in the previous example, the score difference between the target walking scene and its corresponding similar walking scene is 5 points. This score difference value is less than the preset difference value. The difference between the first sensor data and the second sensor data is calculated to obtain the third sensor data, which is then input into the second deep learning model. If the model detects that the target person's walking scene is determined to be the target walking scene, i.e., uphill, then the current walking scene of the target person is determined to be uphill.

[0028] As can be seen from the above, the gait recognition method provided in this embodiment can detect and identify the walking scene of the target person based on sensor data collected by multiple sensors, and accurately identify target walking scenes and similar walking scenes that may have similar gaits, thereby determining the walking scene of the target person. This gait recognition method can solve the problems of low gait recognition accuracy and low efficiency in robots, effectively improving the intelligence level of robots and enabling them to achieve the effect of artificial rehabilitation training.

[0029] In one embodiment of this disclosure, the gait recognition method further includes: If the target walking scene is within the preset walking scene range, it is determined that there are similar walking scenes to the target walking scene.

[0030] In this embodiment, determining that a similar walking scene exists in response to the target walking scene being within a preset walking scene range can be described in detail as follows: A pre-defined walking scenario range is established. When a target walking scenario falls within this range, it can be determined that similar walking scenarios exist. Furthermore, a first mapping relationship is established between the target walking scenario and the similar walking scenarios. This first mapping relationship serves as the basis for determining the similar walking scenarios corresponding to the target walking scenario. Based on the target walking scenario and the first mapping relationship, the similar walking scenarios corresponding to the target walking scenario can be identified.

[0031] This embodiment provides a reference example, such as defining the walking scenario range as uphill, upstairs, downhill, and downstairs; the first mapping relationship can be that when the target walking scenario is uphill, there exists a corresponding similar walking scenario of upstairs; when the target walking scenario is downhill, there exists a corresponding similar walking scenario of downstairs. If, after identification and detection by the first deep learning model, the current target walking scenario is found to be uphill, and this target walking scenario is within the preset walking scenario range, then it can be determined that the target walking scenario has a corresponding similar walking scenario; according to the first mapping relationship, when the target walking scenario is uphill, there exists a corresponding similar walking scenario of upstairs, then it can be determined that the current target walking scenario is uphill, and it has a similar walking scenario of upstairs.

[0032] In one embodiment of this disclosure, the third sensor data includes plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person.

[0033] The data from the third sensor is input into a pre-set second deep learning model to obtain the walking scene of the target person, including: The gait phase is segmented based on plantar pressure data from the third sensor, and the target gait phase is extracted from the first sensor data. The target gait phase is a pre-defined gait phase used to distinguish the target walking scenario from similar walking scenarios.

[0034] The target gait phase data is input into a pre-set second deep learning model to obtain the walking scene of the target person.

[0035] In this embodiment, the third sensor data, obtained based on the first and second sensor data, includes: plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person. The third sensor data is input into a second deep learning model for detection and recognition, which yields the target person's walking scene. It is assumed that the target person has multiple gait phases during walking, and the gait phase that can distinguish the target walking scene from similar walking scenes is designated as the target gait phase. Dividing the third sensor data into gait phases based on plantar pressure data yields multiple different gait phases. Since the plantar pressure changes differently in different gait phases, the plantar pressure data can be used to analyze the changes in the target person's plantar pressure. Based on these changes, the third sensor data is divided to obtain multiple different gait phases. The target gait phase data is extracted from the first sensor data and input into a preset second deep learning model to obtain the determined target person's walking scene.

[0036] This embodiment provides a reference example. For instance, the typical gait phases of a person walking are: foot flat, foot flat - leg flexion, leg swing - continued flexion, leg swing - leg extension, heel off the ground, heel begins to land, and foot lands. Considering that the sensor data for the "leg swing - continued flexion, leg swing - leg extension, etc." gait phases are similar for going upstairs and uphill, to better distinguish between these two walking scenarios, the target gait phases can be set as: foot flat, foot flat - leg flexion, heel off the ground, heel begins to land, and foot lands. The target gait phase data is extracted from the first sensor data and input into the second deep learning model. After feature extraction and detection, the model outputs a walking scenario prediction result of "going upstairs," thus confirming that the target person's walking scenario is "going upstairs."

[0037] In this embodiment, different gait phase division methods can be used not only based on plantar pressure data but also based on time periods. The third sensor data can be divided according to a preset time period to obtain multiple different gait phases. For example, this embodiment sets the preset time period to 2 seconds, dividing the third sensor data into multiple data segments with a 2-second time interval. Based on the analysis of each data segment, multiple different gait phases can be obtained.

[0038] In one embodiment of this disclosure, when the target walking scenario is uphill, upstairs, downhill, or downstairs, the target gait phase includes: foot flattening, foot flattening-leg bending, heel off the ground, heel beginning to fall, and foot landing.

[0039] In this embodiment, the target person may have multiple different gait phases within one gait cycle. This disclosure provides a reference example such as: foot flat, foot flat-leg flexion, leg swing-continued flexion, leg swing-leg extension, heel off the ground, heel beginning to land, and foot landing. Since the sensor data corresponding to the "leg swing-continued flexion" and "leg swing-leg extension" gait phases are similar, it is impossible to determine the corresponding walking scenario using these two gait phases. Therefore, when the target walking scenario is uphill, upstairs, downhill, or downstairs, the target gait phase is set to include: foot flat, foot flat-leg flexion, heel off the ground, heel beginning to land, and foot landing, excluding leg swing-continued flexion and leg swing-leg extension.

[0040] In one embodiment of this disclosure, the gait recognition method further includes: In response to a score difference greater than or equal to a preset difference, the target walking scenario is determined as the walking scenario of the target person.

[0041] In this embodiment, in response to a score difference greater than or equal to a preset difference, the target walking scene is determined as the walking scene of the target person, which can be described in detail as follows: The score difference between the target walking scene and its corresponding similar walking scene is analyzed. When the score difference is greater than or equal to the aforementioned preset difference, the target walking scene is directly determined as the walking scene of the target person.

[0042] This embodiment provides a reference example. For instance, if the preset difference is 10 points, the target walking scenario is uphill, and its score is 85 points, and the similar walking scenario corresponding to the target walking scenario is upstairs, and its score is 70 points, then the score difference between the target walking scenario and its corresponding similar walking scenario is 15 points. This difference is greater than the preset difference, so the target walking scenario is the walking scenario of the target person, that is, the walking scenario of the target person is uphill.

[0043] In one embodiment of this disclosure, the gait recognition method further includes: Data samples from the target walking scenario and similar walking scenarios were selected from sensor data in multiple walking scenarios.

[0044] The second deep learning model is obtained by training the model based on data samples from the target walking scenario and similar walking scenarios.

[0045] In this embodiment, data samples from the target walking scenario and similar walking scenarios are selected from sensor data in multiple walking scenarios. Based on these data samples, a model is trained to obtain a second deep learning model, which can be detailed as follows: Sensor data from multiple walking scenarios are filtered to obtain two data samples corresponding to the target walking scenario and similar walking scenarios, namely the target walking scenario data sample and the similar walking scenario data sample. The key features of the two data samples are extracted using a deep learning model, and the model is trained on the two data samples to obtain a second deep learning model.

[0046] This embodiment provides a reference example, in which target walking scene data samples and similar walking scene data samples are respectively input into the YOLO model. The backbone network of the YOLO model extracts key features from the two samples, and the neck network trains the two data samples to finally obtain the second deep learning model.

[0047] In one embodiment of this disclosure, the sensor includes: Pressure sensor, displacement sensor and inertial measurement unit (IMU).

[0048] The first sensor data includes: plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person.

[0049] In this embodiment, the multiple sensors mounted on the robot include a pressure sensor, a displacement sensor, and an IMU (Inertial Measurement Unit). The first sensor data collected by these sensors includes: plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person. Specifically, the pressure sensor collects the plantar pressure data; the displacement sensor collects the lower limb displacement data; and the IMU collects the lower limb angle, angular velocity, and acceleration data.

[0050] Corresponding to the gait recognition method in the above embodiments, Figure 2 This is a structural block diagram of a gait recognition device according to an embodiment of the present disclosure. For ease of explanation, only the parts relevant to the embodiment of the present disclosure are shown. References Figure 2 The gait recognition device 20 includes: a data input module 21, a score prediction module 22, and a scene determination module 23. The data input module 21 is used to input the first sensor data into a preset first deep learning model to obtain the scores corresponding to each category of walking scenarios. The first sensor data is the data of one gait cycle collected by the sensors set on the robot when the target person is walking; the target person is the person wearing the robot.

[0051] The score prediction module 22 is used to determine the score difference between the target walking scene and similar walking scenes in response to the existence of similar walking scenes in the target walking scene. The target walking scene is the walking scene with the highest score in each category of walking scenes.

[0052] The scene determination module 23 is used to subtract the first sensor data from the second sensor data in response to a score difference less than a preset difference, thereby obtaining third sensor data. The third sensor data is then input into a preset second deep learning model to obtain the walking scene of the target person.

[0053] The second sensor data is the data collected by the sensor when the target person is stationary. The first deep learning model is trained based on sensor data samples in multiple walking scenarios, and the second deep learning model is trained based on sensor data samples in the target walking scenario and similar walking scenarios.

[0054] In one embodiment of this disclosure, the score prediction module 22 is further configured to: If the target walking scene is within the preset walking scene range, it is determined that there are similar walking scenes to the target walking scene.

[0055] In one embodiment of this disclosure, the third sensor data includes plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person.

[0056] In one embodiment of this disclosure, the scene determination module 23 is further configured to: The data from the third sensor is input into a pre-set second deep learning model to obtain the walking scene of the target person, including: The gait phase is segmented based on plantar pressure data from the third sensor, and the target gait phase is extracted from the first sensor data. The target gait phase is a pre-defined gait phase used to distinguish the target walking scenario from similar walking scenarios.

[0057] The target gait phase data is input into a pre-set second deep learning model to obtain the walking scene of the target person.

[0058] In one embodiment of this disclosure, when the target walking scenario is uphill, upstairs, downhill, or downstairs, the target gait phase includes: foot flattening, foot flattening-leg bending, heel off the ground, heel beginning to fall, and foot landing.

[0059] In one embodiment of this disclosure, the scene determination module 23 is further configured to: In response to a score difference greater than or equal to a preset difference, the target walking scenario is determined as the walking scenario of the target person.

[0060] In one embodiment of this disclosure, the scene determination module 23 is further configured to: Data samples from the target walking scenario and similar walking scenarios were selected from sensor data in multiple walking scenarios.

[0061] The second deep learning model is obtained by training the model based on data samples from the target walking scenario and similar walking scenarios.

[0062] In one embodiment of this disclosure, the sensor includes: Pressure sensor, displacement sensor and inertial measurement unit (IMU).

[0063] The first sensor data includes: plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person.

[0064] See Figure 3 , Figure 3 This is a schematic block diagram of a processing terminal provided in one embodiment of this disclosure. Figure 3 The processing terminal 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules / units in the above-described device embodiments, such as... Figure 2 The functions of modules 21 to 23 are shown.

[0065] It should be understood that, in the embodiments of this disclosure, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0066] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0067] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0068] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this disclosure can execute the implementation methods described in the first and second embodiments of the gait recognition method provided in the embodiments of this disclosure, or they can execute the implementation methods of the terminal described in the embodiments of this disclosure, which will not be repeated here.

[0069] In another embodiment of this disclosure, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. The computer program can also instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0070] The computer-readable storage medium can be an internal storage unit of the terminal in any of the foregoing embodiments, such as the terminal's hard disk or memory. The computer-readable storage medium can also be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the terminal. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of the terminal. The computer-readable storage medium is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0071] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0072] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the terminals and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0073] In the several embodiments provided in this application, it should be understood that the disclosed terminals and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0074] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this disclosure, depending on actual needs.

[0075] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0076] The above are merely specific embodiments of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this disclosure, and these modifications or substitutions should all be covered within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A gait recognition method, characterized in that, include: The data from the first sensor is input into the preset first deep learning model to obtain the scores corresponding to each category of walking scene; The first sensor data is data collected by a sensor on the robot during one gait cycle when the target person is walking; the target person is the person wearing the robot. In response to the existence of similar walking scenarios in the target walking scenario, the score difference between the target walking scenario and the similar walking scenario is determined; The target walking scenario is the walking scenario with the highest score among all categories of walking scenarios; In response to the score difference being less than a preset difference, the difference between the first sensor data and the second sensor data is calculated to obtain the third sensor data; The data from the third sensor is input into a preset second deep learning model to obtain the walking scene of the target person; Wherein, the second sensor data is the data collected by the sensor when the target person is stationary, the first deep learning model is trained based on sensor data samples in multiple walking scenarios, and the second deep learning model is trained based on sensor data samples in the target walking scenario and similar walking scenarios.

2. The gait recognition method as described in claim 1, characterized in that, The gait recognition method further includes: If the target walking scene is within a preset walking scene range, it is determined that there are similar walking scenes to the target walking scene.

3. The gait recognition method as described in claim 1, characterized in that, The third sensor data includes plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person; The step of inputting the data from the third sensor into a preset second deep learning model to obtain the walking scene of the target person includes: The third sensor data is divided into gait phases based on plantar pressure data, and the target gait phase data is extracted from the first sensor data; the target gait phase is a pre-set gait phase used to distinguish the target walking scenario from the similar walking scenario; The target gait phase data is input into a preset second deep learning model to obtain the walking scene of the target person.

4. The gait recognition method as described in claim 3, characterized in that, When the target walking scenario is uphill, upstairs, downhill, or downstairs, the target gait phase includes: foot flat, foot flat-leg bend, heel off the ground, heel begins to fall, and foot falls.

5. The gait recognition method as described in claim 1, characterized in that, Also includes: In response to the score difference being greater than or equal to a preset difference, the target walking scene is determined as the walking scene of the target person.

6. The gait recognition method as described in claim 1, characterized in that, Also includes: The target walking scenario and data samples from similar walking scenarios are selected from sensor data in multiple walking scenarios. The second deep learning model is obtained by training the model based on data samples from the target walking scenario and similar walking scenarios.

7. The gait recognition method as described in claim 1, characterized in that, The sensor includes: Pressure sensor, displacement sensor, and inertial measurement unit (IMU); The first sensor data includes: plantar pressure data, lower limb displacement data, lower limb angle data, lower limb angular velocity data, and lower limb acceleration data of the target person.

8. A gait recognition device, characterized in that, include: The data input module is used to input the first sensor data into the preset first deep learning model to obtain the scores corresponding to each category of walking scene; The first sensor data is data collected by a sensor on the robot during one gait cycle when the target person is walking; the target person is the person wearing the robot. The score prediction module is used to determine the score difference between the target walking scene and the similar walking scene in response to the existence of similar walking scenes in the target walking scene; The target walking scenario is the walking scenario with the highest score among all categories of walking scenarios; The scene determination module is used to subtract the first sensor data from the second sensor data to obtain the third sensor data in response to the score difference being less than a preset difference. The data from the third sensor is input into a preset second deep learning model to obtain the walking scene of the target person; Wherein, the second sensor data is the data collected by the sensor when the target person is stationary, the first deep learning model is trained based on sensor data samples in multiple walking scenarios, and the second deep learning model is trained based on sensor data samples in the target walking scenario and similar walking scenarios.

9. A robot, characterized in that, include: A processing terminal and multiple sensors connected to the processing terminal; The processing terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images